Chatterbox TTS API

Minimal FastAPI wrapper around Chatterbox TTS with:

• POST /v1/audio/speech
• GET /v1/models
• GET /health
• startup model loading
• default, Hugging Face repo, and local model loading
• non-streaming WAV responses and SSE streaming
• per-response usage metadata: input_chars and audio duration

Quick Start

git clone https://github.com/travisvn/chatterbox-tts-api
cd chatterbox-tts-api
cp .env.example .env
uv sync
uv run uvicorn app.main:app --host 0.0.0.0 --port 4123

Or with pip:

python -m venv .venv
source .venv/bin/activate
pip install -r requirements.txt
cp .env.example .env
uvicorn app.main:app --host 0.0.0.0 --port 4123

API

Non-streaming

curl -X POST http://localhost:4123/v1/audio/speech \
  -H "Content-Type: application/json" \
  -d '{"input":"Hello from Chatterbox"}' \
  --output speech.wav -i

Response headers include:

• X-Usage-Input-Chars
• X-Usage-Audio-Seconds

SSE streaming

curl -X POST http://localhost:4123/v1/audio/speech \
  -H "Content-Type: application/json" \
  -H "Accept: text/event-stream" \
  -d '{"input":"Hello from Chatterbox","stream_format":"sse"}'

The final speech.audio.done event includes:

• usage.input_chars
• usage.audio_seconds

Models

curl http://localhost:4123/v1/models

Health

curl http://localhost:4123/health

Configuration

Important settings:

• VOICE_SAMPLE_PATH
• DEVICE
• MODEL_SOURCE
• MODEL_CLASS
• MODEL_REPO_ID
• MODEL_LOCAL_PATH
• MODEL_SUPPORTED_LANGUAGES
• DEFAULT_LANGUAGE
• EXAGGERATION
• CFG_WEIGHT
• TEMPERATURE
• MAX_CHUNK_LENGTH
• MAX_TOTAL_LENGTH

See .env.example and .env.example.docker for full examples.

Model Sources

Default bundled loader

MODEL_SOURCE=default
USE_MULTILINGUAL_MODEL=true

Hugging Face repo

MODEL_SOURCE=hf_repo
MODEL_CLASS=multilingual
MODEL_REPO_ID=CoRal-project/roest-v3-chatterbox-500m
MODEL_SUPPORTED_LANGUAGES=da,en
DEFAULT_LANGUAGE=da

Local snapshot directory

MODEL_SOURCE=local_dir
MODEL_CLASS=multilingual
MODEL_LOCAL_PATH=./models/coral-roest-v3
MODEL_SUPPORTED_LANGUAGES=da,en
DEFAULT_LANGUAGE=da

Docker

One maintained GPU-capable Dockerfile is provided in docker/Dockerfile.

Build and run it directly:

docker build -f docker/Dockerfile -t chatterbox-tts-api .
docker run --rm -p 4123:4123 \
  -v "$PWD/voice-sample.mp3:/app/voice-sample.mp3:ro" \
  chatterbox-tts-api

Docs

• docs/API_README.md
• docs/STREAMING_API.md
• docs/MULTILINGUAL.md
• docs/DOCKER_README.md

Tests

python tests/run_tests.py